A two-stage disaggregate attribute choice model is proposed and empirically implemented. The first stage of the model is attribute processing to screen the number of alternatives down to a lesser number. The second stage is brand (alternative) processing which considers the attributes simultaneously while allowing for tradeoffs among the attributes. This two-stage approach is then applied to the same real world data set as two single stage disaggregate models, logit and Maximum-Likelihood-Hierarchical (MLH) which are state of the art models representing the alternative and attribute processing approaches, respectively. The predictive accuracy of the two-stage approach compares favorably to the single stage models. In addition, it seems to offer diagnostic information that can provide managerial insights not found in the output of the single stage model.choice modeling, hierarchical, brand choice, multistage decisions
The authors argue that for the cross-sectional multiattribute approach to choice modeling, the multinomial logit is theoretically and empirically superior to the more commonly used regression approach. Other choice methodologies also are discussed briefly in relation to logit. The difference between individual level (where regression is appropriate) and cross-sectional analysis is recognized. Most marketing managers, because of their research goals, will be using a cross-sectional approach. The derivation of the logit from an underlying behavioral model of choice is illustrated. It is this underlying behavioral model of choice that provides logit with several conceptual advantages for modeling a multiattribute choice structure.
This paper formulates a model for determining the optimal allocation of a given advertising budget over M interacting market segments and a time domain of T periods. Guidance for budget size optimization is provided via a shadow price. The basic input parameters are in terms of sales saturation levels and advertising elasticities, concepts experienced advertising executives understand and are willing to estimate. Economic and political developments in Western Europe are increasing the interactions among geographically defined segments, and recent development in discrimination techniques may soon make market segmentation a practical strategy for the United States market. In relation to the time domain, the simplistic and often unrealistic constraint of a constant carryover effect is relaxed in that current advertising is hypothesized to generate sales both by creating new customers and by interacting with the residual effects of past advertising (goodwill) thus reinforcing preferences of past customers. This interactive formulation relating past and future expenditures allows a model user greater flexibility in specifying the relationship between advertising and sales. The mathematical programming technique used is a form of convex programming.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.